Classification improvement of P300 response based auditory spatial speller brain-computer interface paradigm

The aim of the presented study is to provide a comprehensive test of the EEG evoked response potential (ERP) feature selection techniques for the spatial auditory BCI-speller paradigm, which creates a novel communication option for paralyzed subjects or body-able individuals requiring a direct brain-computer interfacing application. For rigor, the study is conducted with 16 BCI-naive healthy subjects in an experimental setup based on five Japanese hiragana characters in an offline processing mode. In our previous studies the spatial auditory stimuli related P300 responses resulted with encouragingly separable target vs. non-target latencies in averaged responses, yet that finding was not well reproduced in the online BCI single trial based settings. We present the case study indicating that the auditory spatial unimodal paradigm classification accuracy can be enhanced with an AUC based feature selection approach, as far as BCI-naive healthy subjects are concerned.